A remote-controlled, bilateral forcereflecting servo-manipulator has been developed for performing maintenance in cells of nuclear fuel reprocessing plants where human access is not permitted. The in-cell slave unit is constructed from gear trains and torque tubes arranged in modular units so as to be totally maintained or replaced within the cell by another in-cell slave unit. Gear and torque tube construction was chosen for the slave unit because of the modularity feature and because it may be made with good radiation resistance. According to conventional practice, the master controller should be designed so as to be mechanically identical to the slave unit. However, the use of gear trains and torque tubes in the slave unit causes it to have increased inertia, friction, and backlash compared to conventional tape and/or cable driven servo-manipulators. Consequently, if the master controller were constructed of gear trains and torque tubes like the slave unit, the master controller would be so hard to move physically that its operation would quickly exhaust the strength of any operator. Accordingly, simply designing the master controller to be a mechanical duplicate of the slave unit is unacceptable. The master controller must be provided with drive mechanics such as cable and sheave drives, for example, so as to minimize the inertia, friction and backlash, and thereby be easier to operate.
It is not conventional practice, however, to design master controllers that are mechanically different from their slave units. Different mechanical arrangements typically result in the slave unit and the master controller having nonidentical kinematics, the effect of which would be to quickly frustrate an operator's performance.
Geometric identity exists between servomanipulators and master controllers if their link lengths and joint orientations are similar. Kinematic identity includes those aspects, but goes further to require identical force and torque transmission in the servo-manipulator and in the master controller. Kinematic identity means that the entire manner in which forces are transmitted through elements of the slave unit, including joint interrelations such as torsional cross-coupling and positional cross-coupling, is identically replicated in the master controller. In the past, kinematic identity was always assured by mechanical duplication.
A torsionally cross-coupled joint is one in which the torque in that joint is a function of both the load at that joint and the load in another interrelated joint. For example, the summation of torques about the elbow joint of a slave servomanipulator with centralized motors or its master controller is equal to the torque supplied by the elbow motor acting through the elbow joint plus a proportional amount of torque supplied by the wrist motors. Therefore, the torque required from the elbow motor to support the elbow depends on how the wrist is being loaded. If these and other kinematic relationships are not the same in both the slave and master units, the kinematic differences can reduce controllability and create inaccurate force reflection between the units.
If torsional loadings at the joints of the master are not cross-coupled with one another as in the slave, then as the operator moves a member such as the wrist of the master, he would feel the elbow load change. It would get lighter if he moved it one direction, and heavier when moved in the opposite direction. This effect would occur even though the load in the slave elbow was not changing.
The drive mechanism of a slave servomanipulator may also interrelate some joints positionally. Such joints are positionally cross coupled if a change in the position of one joint changes the position of the other.
If, with regard to positional changes, the joints of the master were not coupled to one another the same as in the slave, the master wrist, for example, would not move as the elbow is moved, but the slave wrist would. In such case, the wrists of the slave and master would not move synchronously, which would be unacceptable.
As the state of the art advances, it has been considered to use a master controller having nonreplica kinematics but to modify the control loop of the robotic system to overcome the kinematic differences. The calculations needed for this approach would require very fast software with a very powerful computer in order to have the calculation/control loop close fast enough to make the system stable. Such a control system may be possible in the future. The present invention, however, makes it possible to use a conventional position-position control system such as those in widespread use today even though the master and slave units are mechanically dissimilar.